ESM-IF1
v1.0.0Design protein sequences from 3D backbone structures with controllable sampling diversity.
Design protein sequences from 3D backbone structures with controllable sampling diversity.
ESM-IF1 is an inverse folding model that generates protein sequences from 3D backbone structures. Given a protein backbone, it predicts amino acid sequences that would fold into that shape.
Inverse folding is the opposite of structure prediction. While ESMFold predicts structures from sequences, ESM-IF1 predicts sequences from structures. This enables structure-based protein design - you define the shape you want, and the model suggests sequences to achieve it.
ESM-IF1 was trained on 12 million protein structures predicted by AlphaFold2. It achieves 51% native sequence recovery overall and 72% recovery for buried residues, which are most critical for maintaining structural integrity.
The model uses a GVP-Transformer architecture that combines geometric deep learning with sequence modeling. It processes backbone coordinates through invariant geometric layers, then uses a transformer to generate sequences autoregressively.
The GVP (Geometric Vector Perceptron) layers process backbone atom coordinates in a rotation and translation invariant manner. This means the model understands structural relationships without depending on how the protein is oriented in space.
ESM-IF1 predicts amino acids one at a time, conditioning each prediction on both the full backbone structure and all previously generated amino acids. This autoregressive approach allows the model to maintain sequence coherence while respecting structural constraints.
The temperature parameter controls prediction diversity. Low temperatures (0.1-0.5) produce conservative sequences similar to natural proteins. High temperatures (1.5-2.0) increase diversity, generating more novel sequences that may still fold correctly but differ more from known proteins.
Upload a PDB or mmCIF (.cif) file containing the backbone structure you want to design sequences for. The model uses the N, CA and C backbone atom coordinates to determine amino acid identities.
Generate multiple sequence variants from the same backbone. We recommend generating 8-16 sequences to explore the design space, then selecting candidates based on recovery scores or other criteria.
For multi-chain structures, specify which chain to redesign. The model designs that chain from its backbone coordinates.
Recovery measures what fraction of designed amino acids match the native sequence. Higher recovery means more amino acids match; it does not establish structural compatibility or stability.
Buried residues (inside the protein core) typically show higher recovery than surface residues, since core positions have stronger structural constraints.
We categorize designs by recovery:
These labels summarize sequence recovery. They do not measure experimental stability or folding success.
The mutation list shows positions where the designed sequence differs from native. Review these to understand what changes the model suggests and whether they make biochemical sense.
Stabilize an existing protein by generating variants optimized for the structure. Compare multiple designs and select those with favorable mutations at known weak points.
Create sequence diversity for directed evolution starting points. Generate many variants, then screen experimentally to find improved properties.
Design sequences for computationally generated backbones. Combine with RFdiffusion or other structure generation tools to create entirely new proteins.
ESM-IF1 is optimized for backbones up to 500 amino acids. Longer structures may have degraded performance and slower inference.
The model assumes fixed backbone geometry. It does not account for backbone flexibility or predict how mutations might alter the structure.
Training on AlphaFold2-predicted structures means the model may perform less well on unusual backbone geometries not well-represented in predicted structure databases.
Design protein sequences optimized for thermal stability from backbone structures.
Design protein sequences around ligands, metals, and nucleotides for enzyme engineering and binding-site optimization.
Family-conditioned protein sequence generation from FASTA, A2M, or A3M input
Design amino acid sequences for protein backbones with fixed positions, amino acid biases, and sequence diversity controls.
Design sequences with the ProteinMPNN-family model trained on structures from soluble-protein PDB IDs.
Design sequences for a structure or an aligned ensemble
Design antibody CDR loops from antibody-antigen complex structures using inverse folding.
Joint protein and coding DNA design from protein backbones
Target sequence-conditioned peptide binder design using masked language modeling
Optimize protein binder, cyclic peptide and ligand binder designs with RF3, Boltz-2 and MPNN